Research On The Technology Of Seismic Data Processing Based On Theories Of Blind Signal

To improve the SNR(Signal to Noise Ratio) and increase resolution of seismic data are always two vital processes in seismic data processing. The traditional seismic signal denoising methods are often based on the similarity of effective wave, and at some degree of SNR, which are severe conditions. Thus, it is meaningful to study the noise eliminating way which can break conventional assumption limits. Deconvolution techniques are the main means to enhance seismic prospecting resolution, but the previous ones must determine seismic wavelets in advance. Due to characters of time varying and space transforming in seismic wavelets, it is hard to accurately extract seismic wavelets, so the deconvolution is restricted for application.The recent beginning blind source signal theories are on the strength of basically unknown source information and the lack of priori assumption or restraint. And only depending on analyzing and discussing observed mixed signals, every component of source information can be divided out respectively from blind source. Thereby blind signal separation and blind source deconvolution and so forth have been put forward and these issues can be solved properly.Independent Component Analysis is rather an effective algorithm of blind signal separation methods. On this basis, the paper has studied two aspects of seismic signal blind separation denoising and seismic deconvolution.Firstly, the relevant theory foundation and method categories about blind information decomposition have been briefly introduced in the former part, and Independent Component Analysis has been focused to be analyzed and researched.Secondly, seismic signal with random noise has been simulated as linear transient mixing model, and the desired signal and noise information both have been blindly demixed using FastICA. The uncertainty of results has been analyzed and some appropriate solutions have been explored. Also, in this paper, conditions on different kinds of SNR have been experimented, and influencing factors have been discussed. Besides, an actual stacked seismic data has been processed with blind separation denoising, which has taken a good result.Eventually, considering convolved mixture of seismic records, some time delay transformation has added and the records have become a specific kind of linear transient mixed model in order to meet the requirements of blind signal theories. And blind deconvolution based on ICA has been conducted, while the steps of blind deconvolution and influence factors have been studied and improved. Single channel seismic record and wedge shaped mode seismic records have been made blind deconvolution, and the results mean reflectivity coefficients and seismic wavelets can be separated respectively using FastICA about wavelets of different kinds of phase. This kind of blind deconvolution has a good applying prospect.43pictures,48references.